Abstract: Aiming at the shortcomings of the classic algorithm of UAV inspection image bird’s nest detection on transmission lines,such as excessive weight parameter scale,low recognition efficiency and recognition accuracy,the paper proposes the bird’s nest detection method for an improved YOLOv4 transmission line. Firstly,a mosaic image enhancement method is used to perform various transformations on a picture set to increase the number of small targets in the picture set. Secondly,the depthwise separable convolution is used in the trunk feature extraction network to improve the speed of the detection network. In the YOLO head,the anchor frame size and proportion are improved based on K-means++ algorithm,and the regression loss function is established based on a minimum convex set. Finally,two SPP modules are added between the PANet and the YOLO head to further improve the feature fusion ability and enhance the detection ability of the small targets. A dataset is provided by using the UAV inspection image of a power supply bureau,and the contrasting experiments of the improved algorithm and other target detection algorithms are carried out. The results show that the improved algorithm has higher bird’s nest detection accuracy and lower computing overhead.